International Institute of Information Technology, Hyderabad

About: International Institute of Information Technology, Hyderabad is a education organization based out in Hyderabad, India. It is known for research contribution in the topics: Computer science & Authentication. The organization has 2048 authors who have published 3677 publications receiving 45319 citations. The organization is also known as: IIIT Hyderabad & International Institute of Information Technology (IIIT).

An efficient approach for assessing the seismic soil structure interaction effect for the asymmetrical pile group

Abstract: All of the civil engineering structures involve some type of structural element which is in direct contact with soil. To estimate the accurate response of the superstructure it is necessary to consider the response of the soil supporting structure, and is well explained in the soil structure interaction analysis. Many attempts have been made to model the SSI problem numerically; however the soil nonlinearity, foundation interfaces and boundary conditions make the problem more complex and computationally costlier. To overcome this problem the attempt has been made to optimize the computational efficiency by applying the equivalent pier method for the deep foundation system. In this research paper the L-shape 11 storey building supported by a pile foundation with homogeneous local soil condition is analyzed for dynamic loading including the SSI effect. The significance of the SSI effect has been studied by comparing the responses of the system for fixed base and flexible base condition. A new approach has been proposed to provide simplicity in SSI modeling and reduce the computational cost (both memory and time wise). The approach includes the applicability of the equivalent pier method for the asymmetrical pile groups system, including SSI effect of the pile foundation system. The approach is validated for group effect and found that equivalent pile method can successfully be adopted and helps to reduce the computational cost of SSI problem. To understand the applicability of EPM approach, the parametric study has been carried out for different input of earthquakes and soil types. In accordance with this the three distinct earthquakes, including 1995 Chamba (M = 4.9), 1999 Uttarkashi (M = 6.9) and 2001 Bhuj (M = 7.7) and soil types including cohesive, cohesionless and C-Phi soils have been considered for SSI analysis. The study observed that, earthquake magnitude and soil type shows the major impact on the response of the SSI system.

Discovering Periodic Patterns in Non-uniform Temporal Databases

Abstract: A temporal database is a collection of transactions, ordered by their timestamps. Discovering periodic patterns in temporal databases has numerous applications. However, to the best of our knowledge, no work has considered mining periodic patterns in temporal databases where items have dissimilar support and periodicity, despite that this type of data is very common in real-life. Discovering periodic patterns in such non-uniform temporal databases is challenging. It requires defining (i) an appropriate measure to assess the periodic interestingness of patterns, and (ii) a method to efficiently find all periodic patterns. While a pattern-growth approach can be employed for the second sub-task, the first sub-task has to the best of our knowledge not been addressed. Moreover, how these two tasks are combined has significant implications. In this paper, we address this challenge. We introduce a model to assess the periodic interestingness of patterns in databases having a non-uniform item distribution, which considers that periodic patterns may have different period and minimum number of cyclic repetitions. Moreover, the paper introduces a pattern-growth algorithm to efficiently discover all periodic patterns. Experimental results demonstrate that the proposed algorithm is efficient and the proposed model may be utilized to find prior knowledge about event keywords and their associations in Twitter data.

Noisy Text Data: Achilles' Heel of BERT.

Abstract: Owing to the phenomenal success of BERT on various NLP tasks and benchmark datasets, industry practitioners are actively experimenting with fine-tuning BERT to build NLP applications for solving industry use cases. For most datasets that are used by practitioners to build industrial NLP applications, it is hard to guarantee absence of any noise in the data. While BERT has performed exceedingly well for transferring the learnings from one use case to another, it remains unclear how BERT performs when fine-tuned on noisy text. In this work, we explore the sensitivity of BERT to noise in the data. We work with most commonly occurring noise (spelling mistakes, typos) and show that this results in significant degradation in the performance of BERT. We present experimental results to show that BERT's performance on fundamental NLP tasks like sentiment analysis and textual similarity drops significantly in the presence of (simulated) noise on benchmark datasets viz. IMDB Movie Review, STS-B, SST-2. Further, we identify shortcomings in the existing BERT pipeline that are responsible for this drop in performance. Our findings suggest that practitioners need to be vary of presence of noise in their datasets while fine-tuning BERT to solve industry use cases.

Optimizing Average Precision Using Weakly Supervised Data

Abstract: Many tasks in computer vision, such as action classification and object detection, require us to rank a set of samples according to their relevance to a particular visual category. The performance of such tasks is often measured in terms of the average precision ( ap ). Yet it is common practice to employ the support vector machine ( svm ) classifier, which optimizes a surrogate 0-1 loss. The popularity of svm can be attributed to its empirical performance. Specifically, in fully supervised settings, svm tends to provide similar accuracy to ap-svm , which directly optimizes an ap -based loss. However, we hypothesize that in the significantly more challenging and practically useful setting of weakly supervised learning, it becomes crucial to optimize the right accuracy measure. In order to test this hypothesis, we propose a novel latent ap-svm that minimizes a carefully designed upper bound on the ap -based loss function over weakly supervised samples. Using publicly available datasets, we demonstrate the advantage of our approach over standard loss-based learning frameworks on three challenging problems: action classification, character recognition and object detection.

Semi-Adversarial Networks: Convolutional Autoencoders for Imparting Privacy to Face Images

Abstract: In this paper, we design and evaluate a convolutional autoencoder that perturbs an input face image to impart privacy to a subject. Specifically, the proposed autoencoder transforms an input face image such that the transformed image can be successfully used for face recognition but not for gender classification. In order to train this autoencoder, we propose a novel training scheme, referred to as semi-adversarial training in this work. The training is facilitated by attaching a semi-adversarial module consisting of a pseudo gender classifier and a pseudo face matcher to the autoencoder. The objective function utilized for training this network has three terms: one to ensure that the perturbed image is a realistic face image; another to ensure that the gender attributes of the face are confounded; and a third to ensure that biometric recognition performance due to the perturbed image is not impacted. Extensive experiments confirm the efficacy of the proposed architecture in extending gender privacy to face images.